8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
1/12
Autoresonant beat-wave generation
R. R. Lindberg,a A. E. Charman, and J. S. WurteleUniversity of California, Berkeley Department of Physics, Berkeley, California 94720and Lawrence Berkeley National Laboratory Center for Beam Physics, Berkeley, California
L. FriedlandRacah Institute of Physics, The Hebrew University, Jerusalem 91904, Israel
B. A. Shadwick
Lawrence Berkeley National Laboratory LOASIS Program, Berkeley, Californiaand Institute for Advanced Physics, Conifer, Colorado 80433
Received 7 August 2006; accepted 18 October 2006; published online 7 December 2006
Autoresonance offers an efficient and robust means for the ponderomotive excitation of nonlinear
Langmuir waves by phase-locking of the plasma wave to the slowly chirped beat frequency of the
driving lasers via adiabatic passage through resonance. This mechanism is analyzed for the case of
a cold, relativistic, underdense electron plasma, and its suitability for particle acceleration is
discussed. Compared to traditional approaches, this new autoresonant scheme achieves larger
accelerating electric fields for given laser intensity; the plasma wave excitation is much more robust
to variations in plasma density; it is largely insensitive to the precise choice of chirp rate, provided
only that it is sufficiently slow; and the suitability of the resulting plasma wave for accelerator
applications is, in some respects, superior. As in previous schemes, modulational instabilities of the
ionic background ultimately limit the useful interaction time, but nevertheless peak electric fieldsapproaching the wave-breaking limit seem readily attainable. The total frequency shift required is
only of the order of a few percent of the laser carrier frequency, and might be implemented with
relatively little additional modification to existing systems based on chirped pulse amplification
techniques, or, with somewhat greater technological effort, using a CO2 or other gas laser system.
2006 American Institute of Physics. DOI: 10.1063/1.2390692
I. INTRODUCTION AND OVERVIEW
The plasma beat-wave accelerator PBWA was first pro-posed by Tajima and Dawson TD1 as an alternative to theshort-pulse laser wake-field accelerator, based on earlier
analysis of beat-wave excitation as a mechanism for plasma
heating.24
Subsequently the PBWA concept has been studied
extensively theoretically, numerically, and experimentally514 for a review, see Ref. 15. In the original scheme, twolasers co-propagating in an underdense plasma are detuned
from each other by a frequency shift close to the electron
plasma frequency, so that the modulated envelope resulting
from the beating between the lasers can act ponderomotively
on the plasma electrons to resonantly excite a large-
amplitude, high-phase-velocity Langmuir wave suitable for
particle acceleration.
For a fixed beat frequency, performance of the PBWA is
constrained by what is now known as the Rosenbluth-Liu
RL limit, after the pioneering study in Ref. 4. As theplasma wave grows, relativistic detuning effects eventually
prevent further excitation of the peak longitudinal electric
field Ez beyond a maximum value ERL, which lies below the
cold, nonrelativistic, one-dimensional 1D, wave-breakingWB limit E0:
16,17
ERL E0 163
p2
12
E1E2
E02 1/3 E0 mcpe , 1
where c is the speed of light, m is the electron mass, e is the
magnitude of its charge; p is the electron plasma frequency,
defined in Gaussian units by p2 4n0e2/m, where n0 is theambient electron number density, E1 and E2 are the peak
electric fields of the beating drive lasers, and 1 and 2 are
their respective carrier frequencies. For the plasma waves of
interest here, E0 is smaller than the cold, relativistic or non-
linear wave-breaking limit:15,16 E0EWB =2p 1E0,where p 1 vp
2/c21/2. Here, vp is the phase-velocity ofthe excited plasma wave, approximately equal to the charac-
teristic group velocity vg to be precisely defined below ofthe laser beat envelope, both of which are nearly equal to c.
The detuning effect that gives rise to the limit 1 can beunderstood by considering the effective nonlinear plasma
frequency pNLp/rms, where rms is the root-mean-square relativistic factor of the electrons, including both thetransverse quiver in the laser fields and the longitudinal ve-
locity in the Langmuir wave itself. The transverse quiver
velocity is roughly constant for the fixed-amplitude driving
lasers typical of the PBWA, but as the longitudinal electron
motion in the excited Langmuir wave becomes weakly rela-
tivistic, rms will increase, causing pNL to decrease. Because
of the sensitivity of resonance phenomena, rms need not
grow much before this resonance shift severely limits the
efficacy of the driving beat-wave. As this dynamical processaElectronic address: [email protected]
PHYSICS OF PLASMAS 13, 123103 2006
1070-664X/2006/1312 /123103/12/$23.00 2006 American Institute of Physic13, 123103-1
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
http://dx.doi.org/10.1063/1.2390692http://dx.doi.org/10.1063/1.2390692http://dx.doi.org/10.1063/1.2390692http://dx.doi.org/10.1063/1.2390692http://dx.doi.org/10.1063/1.23906928/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
2/12
is reversible, the Langmuir wave is not actually saturated,
but exhibits a slow nonlinear modulation in amplitude, peri-
odically peaking near ERL as energy is exchanged between
the Langmuir wave and the laser fields.
Tang, Sprangle, and Sudan TSS7 pointed out that onecan, in principle, achieve somewhat higher amplitudes than
ERL by detuning the beat frequency to a value below p,
matching pNL for some nontrivial plasma wave amplitude.
However, for sufficiently large detunings, the high-amplitudesolution cannot be reliably accessed.
18Shvets
19has recently
proposed a complementary approach that takes advantage of
the relativistic bi-stability of the Langmuir wave to excite
large-amplitude waves beyond ERL. Matte et al.20
suggested
countering the change in the resonance using the increase in
plasma density during ionization to compensate for the non-
linear increase in rms; unfortunately, field ionization is ex-
ponentially sensitive to the laser intensity, making this
method difficult to realize experimentally.
As a more practical improvement to,20
Deutsch,
Meerson, and Golub DMG21 suggested incorporating com-pensatory time dependence into the drive laser frequencies.
By chirping the laser frequency downward starting from
the linear resonance p, one can imagine compensating
for the change in the nonlinear resonance with a correspond-
ing change in the beat frequency. DMG invoke the non-
linear dynamical phase-locking phenomenon known as
autoresonance,22,23
arguing that the dynamical system can
self-adjust to maintain phase entrainment between the driven
wave and the ponderomotive drive, resulting in an increase
in the oscillation amplitude without external feedback. How-
ever, proposed as it was before the understanding of au-
toresonance had matured, especially with regard to the
threshold behavior for establishing and maintaining entrain-
ment and the importance of an initial detuning, the DMGscheme sometimes fails to produce appreciable phase-
locking.
Informed by more recent results, we propose a novel
variant of the chirped PBWA concept that also exploits au-
toresonance, but enhanced via adiabatic passage through
resonance APTR.24,25 Rather than chirping downward fromthe linear resonance, we actually start with a frequency shift
well above p, and then slowly sweep the beat frequency
through and below resonance. With this approach, the final
state of the plasma wave is insensitive to the exact chirp
history, and the excitation is more robust with respect to
imprecise characterization of the plasma density. The chirp
rate must only be sufficiently slow so as to satisfy a certainadiabatic trapping condition and thereby ensure phase-
locking. Before the onset of saturation, the frequency of the
excited plasma wave can be closely entrained to the instan-
taneous beat frequency of the drivers, while the plasma wave
amplitude grows monotonically to automatically and self-
consistently adjust itself to the decreasing beat frequency.
Not only is our excitation scheme more robust, so too is
our method of analysis. The Lagrangian fluid formalism de-
veloped by RL and then used by DMG employs a power
series expansion which is valid only for weakly relativistic
motion, so that their equations of motion become increas-
ingly untrustworthy precisely in the regime of interest, where
the plasma wave amplitude grows to some appreciable frac-
tion of E0. Here, we instead employ a fully nonlinear, Eule-
rian fluid model that allows for arbitrarily relativistic elec-
tron motion below the nonlinear wave-breaking limit EWB,
similar to that discussed in Refs. 15 and 26 for laser-plasma
interactions, in Ref. 9 for PBWA investigations, and was first
introduced in Ref. 16.
This analytical fluid model, and the various physical as-
sumptions which go into it, are discussed in Sec. II. In orderto treat the autoresonant nature of the problem more trans-
parently, we reformulate the dynamical equations in a fully
Hamiltonian form in Sec. III. In Sec. IV, we use this canoni-
cal formalism to analyze the autoresonant aspects of the
beat-wave excitation, from the initial phase-locking regime
through the nonlinear trapped phase to nonadiabatic satura-
tion. In Sec. V, we discuss features of some realistic ex-
amples relevant for possible experimental investigation. Sec-
tion VI includes a fluid simulation substantiating our results
with a more complicated laser-plasma model, while Sec. VII
summarizes our preliminary assessment of the merits and
limits of our autoresonant PBWA, especially with respect to
the robust nature of excitation and the utility of phase-locking for matched electron injection. We then offer brief
conclusions from our initial investigation and prospects for
future study in Sec. VIII.
II. FUNDAMENTAL EQUATIONS
Our study of wake excitation is based on an approxi-
mate, but convenient and widely used model. The under-
dense plasma is treated as a cold, collisionless, fully relativ-
istic electron fluid moving in a stationary, neutralizing, ionic
background, coupled to electromagnetic fields governed by
Maxwells equations. The cold, collisionless treatment as-sumes that the electron temperature is sufficiently small so
that the thermal speed is negligible compared to both the
transverse quiver velocity in the driving lasers and the Lang-
muir phase velocity vp c, while stationary ions is valid pro-vided the time scale for ion motion is longer than the dura-
tion T of the lasers; i.e., iT2, where i2 = 4n0Zi
2e2/Miis the ion plasma frequency for ions of mass Mi and charge
+Zie. We restrict our analysis to 1D geometry, assuming the
laser waist and any other transverse variation is much larger
than the Langmuir wavelength p c/p. In addition, weassume prescribed laser fields, neglecting any changes to the
laser envelopes due to linear effects such as diffraction, or
any nonlinear back-action of the plasma on the lasers such asdepletion, self-modulation, or other instabilities.
27,28This
model, although simplified, nevertheless reveals the essential
features of autoresonance and its potential advantages for the
PBWA.
We define a scaled, or dimensionless, time pt,co-moving position ptz/vg, vector potentialae/mc2A, and electrostatic potential e/mc2.Now, we further make the quasi-static approximation QSAsee, e.g., Refs. 9 and 29, wherein we assume that theplasma response is independent of time in the co-moving
frame, i.e., =, and the plasma wave moves withoutdispersion at the fixed average group velocity vg of the driv-
123103-2 Lindberg et al. Phys. Plasmas 13, 123103 2006
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
3/12
ing lasers. Thus, in the QSA, all derivatives in the fluid
equations are neglected, and the continuity and momentum
equation can be explicitly integrated and solved, yielding a
single second-order differential equation for the electrostatic
potential. In the high phase-velocity limit appropriate to un-
derdense plasmas, this equation may be simplified to
2
2=
1
2
1 + a2
1 + 2
1
. 2
After solving for in 2, we can determine the electricfield using /=pEz/E0. While 2 remains math-ematically well defined for any 1, it can only be physi-
cally trusted for electric fields less than EWB.
In the Coulomb gauge, the vector potential is solenoidal,
which in 1D implies that it is geometrically transverse,
so a=a, will be polarized perpendicular to z. Fortwo slowly chirped, nearly-plane-wave, flat-topped lasers
with positive helicity, the normalized vector potential repre-
senting the beating lasers may be written as a=a1 +a2=
1
2e+a1e
i1 + e+a2ei2 +c.c., where e = 1 /2x iy. At
the leading edge of the plasma, the laser phases are given
by j0 , t =j0 0tdtjt, where the j0 depend on initial
conditions, and the instantaneous carrier frequencies
jt j0 , t d/dtj0 , t allow for slow chirping ofone or both of the lasers, such that j
1d/dtjpjand 1 2 p. We define the average carrier frequenciesj 1 /T0
Tdtjt, and the overall average carrier fre-quency by =
1
21 +2.
We assume that the laser frequencies j = /tjand wavenumbers kj = /zj each satisfy the 1D electro-magnetic dispersion relation j
2 =p2/0 + c
2kj2, where the
constant 01 parametrizes a shift in the effective plasma
frequency due to transverse electron quiver. For electro-
magnetic waves satisfying this dispersion relation, the groupvelocity is approximately given by vgd/dkk cc/ 20p
2/2. This expression evaluated at the average
carrier frequency defines our reference group velocity vg:
vg vg c1 120
p2
2 2 1
k2 k1, 3
which represents the characteristic velocity at which both
laser envelope modulations and Langmuir phase-fronts
travel. Using the assumed vector potential, the ponderomo-
tive force is given by
a2 =1
2a
1
2 + a2
2 + a1
*a
2
ei21 + a1
a2
*ei21. 4
Since the group-velocity dispersion effects remain small in
the underdense regime, linear propagation into the plasma
results in a beat phase given by
2z,t 1z,t = 0 + 0
tz/vg
dt2t 1t
+ O 10
p3
3pL
c , 5
where 0 is equal to the difference of the initial laser
phases. The neglected terms limit the validity of the constant
group velocity approximation to interactions lengths L less
than the so-called dispersion length Ldisp 3/p
3c/p.For accelerator applications, the useful interaction length is
already limited by the dephasing length Ld, beyond which
accelerated electrons cannot gain energy from the electro-
static field: LLd 2/p
2c/pLdisp, so that our con-stant group velocity approximation imposes no further re-
striction on the interaction length.
By appropriate choice of the initial phases, we maytake both a1 and a2 to be real. Defining the co-moving nor-
malized beat frequency =2 1/p, the beatphase =0 + 0
d, the normalized beat am-plitude = a1a2, the average normalized intensity per laser
a2 =1
2a1
2 + a22, and the electron quiver factor 0 =1 + a2, the
slow part of the ponderomotive drive may be written as a
function of only:
a2 = a2 a2 + cos .
Using this form for the ponderomotive drive, the equation of
motion 2 may be written as an ordinary differential equa-tion in the co-moving coordinate :
d2
d2= =
1
21 + a2 + cos
1 + 2 1 , 6
with the initial conditions 0 =0 =0. Equation 6 isused for all numerical simulations discussed in Secs. IV, V,
and VII, and is the starting point for our analysis of autoreso-
nance.
III. HAMILTONIAN FORMALISM
To study autoresonance, we now develop the Hamil-
tonian formulation of 6. Our goal is an expression in termsof canonical action-angle variables, for which the phase-
locking phenomenon is most readily analyzed. First, note
that the dynamical equation for the electrostatic potential 6can be derived from the Hamiltonian
H,p; =1
2p2 +
1
2 1
1 + + 1 + a2 + cos
21 +
H0,p +a2 + cos
21 + , 7
with the scalar potential regarded as the generalized coor-
dinate, p d/d=pEz/E0 regarded as the canonicalmomentum conjugate to , and taken as the time-like
evolution variable. In this way, the plasma-wave dynamicsare seen to be analogous to those of a one-dimensional
forced nonlinear oscillator. The component H0 of 7 repre-sents the Hamiltonian of the free oscillator, involving one
term Tp = 12p2 analogous to the kinetic energy of the oscil-
lator which is proportional to the electrostatic energy den-sity of the plasma wave, and another term corresponding toan anharmonic effective potential V = 1
21 /1 + + 1.
The remaining driving term H,p ; H0,p corre-sponds to the time-dependent forcing of the oscillator.
In the absence of forcing i.e., = a2 = 0, the dynamicsgoverned by H0,p =Tp +V are conservative, so theoscillator energy, i.e., the value H of the Hamiltonian H0
123103-3 Autoresonant beat-wave generation Phys. Plasmas 13, 123103 2006
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
4/12
along any particular unperturbed orbit, remains constant. In
the physically allowed region 1, the effective potential
V is non-negative and possesses a single minimumV0 =0, while V as 1+ or . So for anyvalue of energy H0, there exists a phase space trajectory
;H ,p;H which is a closed periodic orbit.Making a canonical transformation to the action-angle
variables of the free oscillator, =I,, p =pI,, we
can express 7 as
HI,; =H0I +a2 + cos
21 + I,, 8
where the action I is defined in terms the area in phase space
contained within the unperturbed orbit ;H ,;H ofenergy H:
I1
2 pd= 1
+
d. 9
Here, + and are the upper and lower turning points of
the orbit, respectively, given by
=HH2 + 2H, and wehave used symmetry to reduce the integration path in 9 tothe segment where p0. By making the change of variables
=+ + sin2u, the action 9 can be calculated us-
ing a standard integral table:30
I=2+
2
1 + +
0
/2
dusin2ucos2u
1 + 1 + +
sin2u
=4
31 + H H2 + 2H1/2
1 + HE
1 + H H2 + 2H K. 10
Here, K and E are complete elliptic integrals of thefirst and second kind, respectively, whose modulus satisfies
2 =21 +HH2 + 2H 2H2 +H. At this point, one couldin principle use 10 to find H0I, but fortunately, such acumbersome inversion will not be necessary. The normalized
i.e., dimensionless nonlinear frequency H of the un-forced oscillator is given by
H = IH
1 = 2
1 + H H2 + 2H1/2
E. 11
To put 7 in the desired form 8, the remaining ingredientneeded is the canonical transformation =I,. An ex-plicit formulation of this will also not be needed, and we
may proceed by formally assuming that we have made this
substitution. Then, =I, is a periodic function of, andwe can expand the driving term appearing in 8 in a Fourierseries as
a2 + cos
21 + I,= a2 + cos
n=
bnIein. 12
Because 12 is a real-valued function, the Fourier coeffi-cients must satisfy bn = bn
* , and are defined by
bnI =1
2
0
2
dein
21 + I,. 13
We put 13 into a form more amenable to calculation bychanging variables using =H, and integrating over oneperiod 0 2/ of the co-moving coordinate. Fur-thermore, if we choose the origin of the orbit associated with
the energy H, such that = 0 ;H =+
, the potential
;H, and hence 1 +;H1, are symmetric about thepoint =/2. Since the imaginary parts of 13 are obtainedby integrating over the antisymmetric functions sinn,the bn values are purely real and are given by
bnH = bnH* =
1
H
0
H
dcosnH
21 + ;H.
Up to this point, no approximations have been made in
the canonical transformations to the action-angle variables of
the unforced oscillator. Now, we make the single resonance
approximation31 SRA to 8, by assuming that the rapidly
oscillating terms of the Hamiltonian average to zero and con-tribute negligibly to the dynamics. Anticipating the develop-
ments of Sec. IV, we realize that under certain frequency-
locking conditions derived there, autoresonant excitation
occurs, wherein the plasma wave amplitude, or equivalently,
the free-oscillator energy H =T()+V(), growssecularly such that the dynamical frequency of theforced oscillator approximately matches that of the unforced
oscillator at that same amplitude, (H). Simulta-neously, this instantaneous oscillator frequency follows that
of the driving beat frequency of the wave, .Under these assumptions, we can consistently neglect all
terms in the sum 12 except the constant term or those terms
with frequency dependence . The aver-aged Hamiltonian then becomes
HI,; =H0I + b1Icos + a2b0I.
We now make an explicitly -dependent canonical trans-
formation to the rotating action-angle variables I,.
Using the mixed-variable generating function F2I,;
= I, our old and new coordinates are related by
=F2/I= and I=F2/=I. Dropping the caret
from the new action, the Hamiltonian in the rotating frame is
KI,; = H0I I+ b1Icos + a
2
b0I .
The resulting SRA canonical equations of motion are
d
d =I +
b1
Icos + a2
b0
I, 14a
d
dI= b1Isin , 14b
for which we next determine the required conditions for
phase-locking.
123103-4 Lindberg et al. Phys. Plasmas 13, 123103 2006
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
5/12
IV. AUTORESONANT RESPONSE
The essential ingredients for autoresonance are: a non-
linear, oscillatory degree of freedom in our case, the plasmawave evolving, in the absence of any forcing, within anintegrable region of phase space; a continuous functional re-
lationship between the nonlinear frequency and energy of the
oscillation possessing a well-defined linear limit; an applied
oscillatory driving force in our case, the modulated pon-deromotive envelope of the lasers which can be consideredperturbative, so that the notion ofthe nonlinear frequency of
the unforced oscillator remains meaningful; and initial con-
ditions and forcing profile consistent with adiabatic passage
through resonance APTRnamely, an initially unexcitedsystem quiescent plasma, and an initial drive frequencyappropriately far from the linear resonance, with subsequent
time dependence that is sufficiently slowly varying but
otherwise arbitrary.
The key consequence of autoresonant beat-wave genera-
tion is the robust entrainment between the three relevant fre-
quencies: the beat frequency of the driving lasers, the
instantaneous frequency of the driven plasma wave,and the nonlinear frequency H of the unforced plasmawave. Assuming this phase-locking is achieved, amplitude
control of the plasma wave can be simply understood via
11: because the frequency of the freely evolving nonlinearplasma wave is a function of the energy, phase-locking of the
driven wave to the envelope such that H implies that changing the drive frequency will cor-respondingly change the oscillator energy. In our case, 11indicates that H is a decreasing function of H, so that inorder to increase the plasma wave amplitude, one must de-
crease the beat frequency as a function of .
While we have indicated how autoresonance can lead to
large plasma waves, we have not yet shown under what con-ditions such phase-locking occurs. To address this question,
we first consider the linear and weakly nonlinear response,
valid for moderate Langmuir amplitudes. Next, we consider
the fully nonlinear case, for which we derive adiabaticity
requirements for autoresonance.
A. Small-amplitude response and phase-locking
When the drive is first applied with its frequency above
the linear resonance, the plasma wave amplitude and, hence,H and IH are small and we can linearize Eqs. 14a and14b. In this limit, the oscillator is harmonic, with
H = 1, =2Hcos, and H=I, so that b0 = 12H,b1 =2H/4, and Eqs. 14 become
d
d = 1
1
2a2 +
421
Icos, 15a
d
dI=
22Isin. 15b
Note that the 1 12
a2 contribution to d/d corresponds,in normalized units, to the leading-order expansion of the
effective plasma frequency in the electromagnetic dispersion
relation
peff p/0 = p/1 + a2 p1 1
2a2 + ,
which is shifted from the bare value p due to the transverse
quiver motion of the electrons in the applied laser fields.
Physically, this implies that one must ensure that the drive
frequency begins sufficiently far above, and then is slowlyswept past, the effective frequency peff, rather than the bare
frequency p. One may wonder why only the leading-order
correction appears here, while we never explicitly invoked
any small-a2 approximation. This is because when making
the SRA, we ignored terms of the form a2ein for n1,while such terms can appreciable effect the dynamics for
large intensity a2 despite being off-resonance.
For the driven plasma wave, we have 1,while we seek solutions for which 1 a s aresult of special initial and forcing conditions: an initially
unperturbed plasma, = 0 == 0 =0; an initial tuning
of the beat frequency above resonance, i.e., = 0peff, and subsequently a slow downward frequency chirpthrough resonance, where the chirp rate is be characterized
by a parameter d/d, with 01.For a linear frequency chirp around the effective plasma
frequency,
= 1 12 a2 , 16
the simple harmonic oscillator equations 15 have analytic
solutions in terms of the Fresnel sine and cosine integrals. 32
We briefly summarize the extensive characterization of these
solutions found in Ref. 24. When the drive is first applied far
from resonance, the oscillator response can be divided into
two components: one ringing component precisely at peff,
and the other at the driving frequency 0, both of smallamplitude. The singular term I1/2 in 15a allows for alarge change in phase at small amplitude without violating
the requirement that remain smooth, so that the re-sponse at the driven frequency can adjust itself to the drive,
and phase-locking can occur. As the frequency is swept to-
ward the resonance, these driven, phase-locked oscillations
grow. Meanwhile, the response at the resonant frequency hasno such phase relation with the drive, and remains small. In
this way, we essentially have one growing, phase-locked
plasma wave when the drive frequency reaches the resonant
frequency.
As the resonance is approached, the amplitude of the
plasma wave begins to become large and one must account
for the growing nonlinearities. We therefore expand the ex-
pression for the free nonlinear frequency 11 to first order:H = 1 3
8H= 1
3
8I, and continue to use the linearized
frequency chirp 16. Making the change of variableA 42I, we have
123103-5 Autoresonant beat-wave generation Phys. Plasmas 13, 123103 2006
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
6/12
d
d =
3
256A
2 +
Acos,
17d
dA = sin.
This set of equations can be reduced to a single first-order
ordinary differential equation by defining the complex
dynamical variable Z 256/3Aei, a re-scaled indepen-dent variable , and the dimensionless parameter3 /2563/2, to obtain
id
dZ+ Z2Z= . 18
Thus, the weakly nonlinear problem is now described by a
dynamical equation with a single parameter that combines
the drive strength and the chirp rate . It has been found
numerically33
that the solution to 18 has a bifurcation at thecritical value =c 0.411. For c, the plasma waveresponse quickly dephases from the drive, resulting in only
small excitations. In contrast, for c, phase-locking oc-curs and the plasma wave can grow to large amplitude. This
critical behavior with respect to translates into a critical
drive strength and chirp rate for the nonlinear oscillator
to be autoresonantly excited. The condition is
32256c
22/3 0.1694/3 . 19Thus, for a given laser intensity, one can readily find the
maximum chirp rate that can be tolerated and still obtain
high-amplitude plasma waves. We demonstrate the sensitiv-
ity of this critical behavior for three drive strengths in Fig. 1,
which plots numerical results from the full quasi-static equa-
tion of motion 6. Plot a shows the dependence of thechirp rate of the maximum amplitude about the critical chirp
rate c. Plot b demonstrates the dynamic frequency-lockingthat occurs in autoresonance for =0.005, and compares this
to the case in which autoresonance fails to occur.
B. Fully nonlinear autoresonant response
If phase-locking is maintained through the weakly non-
linear regime, the amplitude continues to grow and one must
consider further nonlinearities beyond those included in 18.
In this case, there arises more stringent restrictions on thechirp rate for adiabatic phase-locking to persist. We calculate
this condition by first finding a second-order equation for the
phase . Taking the time derivative of 14a, the phase isseen to obey the following second-order equation:
0 =d2
d2 +
b1
Isin
d
d +
d
d
b1IsinI
+ a2
2b0
I2
+
2b1
I2
cos . 20Now, we assume see, e.g., Ref. 34 that the free action canbe written as
I= I0 + I, 21
where I0 =I0 is the slowly varying, secularly evolvingaction about which there are small oscillations given by
I=I. These oscillations correspond to fluctuations in
about its slowly varying phase-locked value =,an example of which can be seen in Fig. 1b. In the au-toresonant case solid line, we see that as the plasma wave is
excited, its frequency does indeed make small oscillationsabout the drive frequency.
Using the form 21 for the action, the lowest-orderequation for the phase is identical to 20, with I being re-placed everywhere by I0. In this way, the phase itself is seen
to obey a nonlinear oscillator equation, with an effective
nonlinear damping term, and a conservative forcing, de-
scribed by the terms on the second line, which is derivable
from an effective potential whose shape is dictated by the
slowly evolving action I0 and the drive parameters and .
Phase-locking then corresponds to trapping of in a basin
of this effective potential about , such that the nonconser-
vative term on the first line must either provide damping or
FIG. 1. Demonstration of the critical autoresonant behavior. In a, we seethat chirp rates slightly below c grow above linear wave-breaking, while
chirps a tiny fraction too large saturate below E0. b compares the phaselocking in autoresonance solid line to that of nonautoresonant behaviordashed line for =0.005. The critical drive mixed line is included forreference.
123103-6 Lindberg et al. Phys. Plasmas 13, 123103 2006
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
7/12
remain small if excitatory. Typically, the ratio of this noncon-
servative term to the second term in the conservative force is
given by
1b1
b1
I0
I0 d
d 1
b1
b1
d
d d
d 1
because the nonlinear phase oscillations are slow compared
to p
.
Actually, over most of the typical range of parameter
values, the last two terms in the conservative forcing
are small compared to the first two terms, and one can
understand the dynamics of the phase oscillations using
the biased pendulum equation d2/d2+b1I0/I0sin, although we will continue to workwith the full equation 20. Clearly, for a given I0, if thenormalized chirp rate remains sufficiently small compared
to the normalized drive strength , then the effective poten-
tial will be of the tilted-washboard variety, with a series of
periodically spaced local minima in at intervals of 2. As
increases, the depth of these wells decreases, until they
finally disappear, as does any opportunity for phase-locking.Thus, a necessary condition for trapping is that the ef-
fective force can actually vanish at some fixed point :
+ b1I0sin I0
+ a2
2b0
I02
+
2b1
I02
cos = 0 .22
For given I0, this determines the average, slowly varying
phase about which trapping occurs. In the linear
I00, weakly forced 01 or weakly chirped
0+ regimes, this phase value is known24 to be ,but for nonlinear plasma waves, stronger forcing, or faster
chirping, this phase can shift appreciably. Since sin ,
cos1, Eq. 22 also imposes an upper bound on thefrequency chirp for which the phase can remain trappedin autoresonance, regardless of the actual value of the phase
at which locking occurs. Setting sin = cos =1 above,we obtain an upper bound on beyond which any phase-
locking is impossible:
b1I0 I0
+ a2 2b0I0
2 + 2b1I0
2 . 23Again, for realistic parameters, the force balance typi-
cally resides predominately between the first two terms in22, and hence this upper bound, although approximate, isexpected to provide a reasonably tight cutoff for autoreso-
nant phase-locking, which has been confirmed by numerical
simulation. This inequality can also be thought of as giving
the maximum achievable plasma wave amplitude implicitlyas a function of I0 for a given chirp rate and laser power.We show the dependence of the saturated longitudinal field
on the chirp rate as a solid line in Fig. 2a for a number ofdifferent drive strengths. For a fixed drive strength , the
maximum attainable electric field jumps discontinuously at
the critical chirp rate c given by 19. The dotted lines cor-respond to solutions of 23 that cannot be accessed when
starting from vanishing initial longitudinal field, and as such
constitutes a form of hysteresis in the excitation; the solid
lines show the stable branches for the case of interest. Figure
2b shows a comparison of the theoretical maximum ampli-tude and that found by numerically integrating the quasi-
static equation of motion 6 for =0.005.Asymptotic expansions and numerical plots for I
and bnI reveal that the right-hand-side of 23 decreaseswith increasing I0 beyond moderate values ofI0. Thus, the
adiabatic nonlinear phase-locking condition 23 become in-creasingly stringent, and the growing plasma wave eventu-
ally falls out of frequency-locking. Mathematically, this satu-ration can be traced to the nonlinear nature of the forcing in
6, where the effective strength depends both on the externaldrive intensity and the plasma wave amplitude; physically, to
the fact that as the plasma wave grows, the electrons become
increasingly relativistic and therefore less susceptible to dis-
placement via the ponderomotive forcing.
As a result of relativistic detuning, recall that in the
original RL/TD scheme, the plasma wave amplitude exhibits
slow compared to p1 nonlinear modulations that appear as
beating; i.e., periodic amplitude oscillations up to the RL
limit and back to a nearly unexcited state. As the wave is
driven to a high amplitude, its phase slowly slips until it is
FIG. 2. Shows the maximum plasma wave amplitude obtainable before the
slowness condition 23 is violated. In a, we plot the maximum longitu-dinal field as a function of the chirp rate for four different driving strengths.
Dotted lines indicate hysteresis at c given by 19. Panel b compares thetheory line with numerically determined saturation for =0.005.
123103-7 Autoresonant beat-wave generation Phys. Plasmas 13, 123103 2006
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
8/12
more than /2 out of phase with respect to the laser beatand then gives its energy back to the lasers, then continues to
shift further out of phase, only to be re-excited when its
amplitude approaches zero and it can re-establish phase
matching with the drive.
In the DMG scheme, the plasma wave amplitude not
only can peak at a higher maximum than in the original
scheme, but typically will sustain a higher average value at
long times, exhibiting a nonlinear ringing about some non-zero saturated value. In this case, when the frequency lock-
ing fails the wave phase is closer to its neutral value with
respect to energy exchange with the drive, and the frequency
difference and phase lag then grow secularly in time. De-
pending on initial parameters, the growth to the absolute
maximum can either be essentially monotonic, or exhibit in-
termittent plateaus or dips between periods of resumed
growth, before finally leveling off.
In our autoresonant scheme based on APTR, the behav-
ior more closely resembles that in DMG scheme, but exhibits
more nearly monotonic growth, higher peak fields, and less
ringing after saturation. The extent of the amplitude and
phase excursions is determined by how deeply the phase istrapped in its effective potential well. This in turn depends
both on how steep and how deep is the available well de-termined by the Langmuir amplitude and drive lasers param-
eters, and to what extent the phase can be nudged near thebottom of the well and kept there determined by the initialconditions and the adiabaticity of the chirped forcing. Nu-merical simulations suggest that the depth of this trapping is
improved by using a stronger drive, starting the drive fre-
quency further above resonance, and chirping more slowly.
In practice, of course, each of these strategies involves trade-
offs. Increasing the drive strength increases the growth rates
for laser-plasma instabilities that might disrupt the forcing.
Either increasing the initial frequency up-shift or decreasing
the chirp rate decreases the final amplitude that can be
reached during a fixed interaction time.
V. EXPERIMENTAL CONSIDERATIONS
Unfortunately, as has been alluded to previously, the
PBWA does not have unlimited time to be excited, as delete-
rious instabilities will eventually destroy wave coherence.
For the parameters of interest, the oscillating two-stream
also referred to as modulational instability limits the life-time of coherent Langmuir waves to the ion time scale, i.e.,
for times of the order of a few 1/i. Although it is possiblethat the growth of this instability may be mitigated somewhat
by the use of a chirped laser, in this paper, we will use as a
conservative figure the results of Mora et al.11
to set the time
limit during which we can excite a coherent plasma wave
suitable for accelerator applications.
For the relatively cold plasmas and moderately intense
lasers we consider, it is shown in Ref. 11 that the growth rate
of the oscillating two-stream instability is approximately
equal to i, and that this instability impedes plasma wave
excitation and destroys coherence after about five e-foldings.
Thus, we see that the drive lasers should have time duration
T5 /i. If one chirps the drive frequency leading to a total
shift during the autoresonant excitation, then the normal-
ized chirp rate is limited to
0.2p
i
p, 24
or approximately 2.3103/p for singly ionized
helium. Below, we choose two experimentally relevant
parameter sets, one corresponding to a 10 m CO2 laser;
the other, to a 800 nm chirped pulse amplification35 CPATi:sapphire laser system. We demonstrate how, beginning
with the laser frequency above the linear resonance and then
slowly decreasing it, one can robustly excite plasma waves
to amplitudes larger than the cold, linear wave-breaking limit
in times commensurate with onset of the oscillating two-
stream instability.
A. CO2 Laser at 10 m
We first consider parameters roughly corresponding to
the most recently published UCLA upgrade.14
We assume
two pulses of duration T=100 ps that enter the plasma at
t=0, whose central wavelengths of 10.27 and 10.59 m im-
ply a resonant plasma density n0 11016 cm3. The lasers
have normalized intensities a1 = a2 =0.14, corresponding to a
normalized drive strength = 0.02, so the threshold condition
19 implies that the normalized chirp rate should satisfyt = d/dtt/p0.0009. We choose a linear chirp,so that in physical units the beat frequency is given by
t = 0 +pt, where 0 =1.15 and =0.000 65, with atotal frequency sweep from t= 0 =1.15 to t= T= 0.74. For these parameters, we collect the results from
simulations integratingthe quasi-static equation of motion 6in Fig. 3. Figure 3a demonstrates the excitation of a uni-
form accelerating field Ez of 10 GV/m, which is above thelinear, cold, wave-breaking limit of E0 8.8 GV/m, but be-low the cold relativistic limit EWB 61 GV/m. The totalchirp is modest, only about 1.5% of the laser carrier fre-
quency 2c/.
For comparison, we also plot the simulated envelopes of
the longitudinal field for the resonant RL/TD case t=0 =1, and for the chirped DMG scheme starting onlinear resonance, t =1 pt, but using the same chirprate as above. The resonant case demonstrates the character-
istic RL limit of EzERL= 16/ 31/3E0 4.2 GV/m,
whereas the DMG scheme fails to achieve appreciable dy-
namic phase-locking, and the final plasma wave amplitude is
about the same as in the resonant unchirped case. Usingapproximately these parameters, UCLA experiments have in-
ferred accelerating amplitudes up to 2.8 GV/m13
over short
regions of plasma. More recently, plasma density variations
corresponding to Ez 0.20.4 GV/m have been directlymeasured with Thomson scattering.
36
Perhaps more important than the higher-amplitude field
in the autoresonant APTR case, is the fact that excitation is
very robust with respect to mismatches between the beat fre-
quency and the plasma frequency. In practice, these mis-
matches inevitably result from limited diagnostic accuracy or
shot-to-shot jitter in the plasma or laser parameters. Because
one sweeps over a reasonably broad frequency range and one
123103-8 Lindberg et al. Phys. Plasmas 13, 123103 2006
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
9/12
only needs to pass through the resonance at some indetermi-
nate point during the chirp history, no precise matching is
required, and the exact value of the plasma density need not
be accurately known. This robustness is demonstrated in
Figs. 3b and 3c. Plot b shows the longitudinal field pro-file attained when there is a 10% variation in the density,
demonstrating that APTR yields uniform, large-amplitude
fields, while forcing the plasma wave off-resonance yields
small accelerating gradients. In Fig. 3c, we plot the finalaccelerating gradient achieved via APTR when we vary the
value ofp over a range of 10%, from its design value,
while keeping the laser parameters fixed. We see large levels
of excitation for a wide range in plasma variation, corre-
sponding roughly to density mismatch/errors up to 20%. On
the other hand, the traditional PBWA scheme only signifi-
cantly excites the Langmuir waves for plasma densities suchthat the resonance condition is nearly met. Thus, not only is
autoresonant plasma wave excitation effective in avoiding
saturation from detuning, it also mitigates experimental un-
certainties in or shot-to-shot variations of plasma density.
B. Ti:sapphire laser at 800 nm
In this section, we analyze a representative case for a
Ti:sapphire CPA laser in a singly ionized He plasma, with
n0 =1.41018 cm3, so that p/=1/25, and a laser dura-
tion T=3.2 ps, chosen to correspond to the modulational in-
stability limit. If we consider two 2-J pulses compressed to
this time duration and focused to a waist of w0 30 m, thisimplies intensities of I0 =2.01017 W/cm2, so that with the
laser wavelength 800 nm, we have a1 = a2 =0.3, and=0.09. We choose t= 0 =1.2, t= T =0.5, for a nor-malized chirp =0.0025. The resulting plasma wave excita-
tion is shown in Fig. 4a. Here, we see maximum longitu-dinal electric fields Ez 260 GV/ m, corresponding to1.6E0 0.25EWB. For comparison, we also plot the reso-nant case, for which detuning results in maximum fields
corresponding to the familiar RL limit 16/ 31/3E0 125 GV/ m, and the DMG case starting on-resonance,which yields results similar to the APTR case. The distinc-
tion between passing through resonance and starting on-
resonance can be seen, however, in Fig. 4b, for whichDMG scheme starting on-resonance does not experience the
fortuitous frequency-locking, as in case a. In contrast, Fig.4b shows that the uniform accelerating fields obtained viaAPTR are only slightly affected by the change in the reso-
nant plasma frequency. Finally, Fig. 4c shows the robust-ness of autoresonant excitation, for which density imperfec-
tions of 35% have little effect on the accelerating gradients
achieved.
VI. FLUID SIMULATIONS
In order to apply the formalism developed for autoreso-
nant excitation, we have made a series of simplifying as-sumptions to arrive at the driven, nonlinear ordinary differ-
ential equation 6. Namely, we have assumed 1D dynamics,quasi-static plasma response, and neglected laser evolution.
To test the latter two assumptions and to validate some of our
conclusions using a more faithful laser-plasma model, we
have simulated the experimental parameters of Sec. V using
a fully relativistic, Maxwell-fluid code in 1D.37
This code
solves the cold electron fluid equations coupled to Maxwells
equations in a frame moving at the speed of light c.
The accelerating electric fields obtained in the fluid
simulations for the resonant RL/TD PBWA, the DMG chirp
from resonance, and the APTR chirp are included in Fig. 5.
FIG. 3. Plasma wave excitation for a 10 m CO2 laser with intensity 2.7
1014 W/cm2 =0.02. In a, the APTR case has 0 =1.15p, and=p at t40 ps =0.000 65. Total chirp is 1.5% of laser frequency. Forcomparison, we include the envelopes with no chirp, and DMG chirp start-
ing on resonance. Panel b is the same as a, with p changed by 10%.Panel c demonstrates robust field excitations of 510 GV/m for densityerrors of order 20% for APTR, while the traditional PBWA only effectively
excites Ez near resonance.
123103-9 Autoresonant beat-wave generation Phys. Plasmas 13, 123103 2006
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
10/12
Note how the qualitative behavior of the excitation is very
close to the QSA results in Figs. 4a and 4b. The solidline in Fig. 5a demonstrates the Rosenbluth-Liu limitERL 125 GV/ m characteristic of resonant driving, whilethe DMG chirp from resonance fails to adequately phase
lock the laser drive to the plasma wave, causing premature
saturation. Figure 5b, on the other hand, demonstrates large
amplitude excitation to Ez 200 GV/m, above the cold,wave-breaking field of E0 160 GV/m, but a bit less thanthe predicted 260 GV/m. The laser drive envelope has un-
dergone some nontrivial evolution, but retains good
frequency-locking to the plasma wave throughout the
window.
VII. DISCUSSION: COMPARISONS, SCALINGS,AND PHASE LOCKING
In comparison to other PBWA schemes, including the
fixed beat-frequency approach at linear resonance RL/TD,the chirped DMG scheme, involving downward chirpingfrom resonance, or the nonresonant PBWA, scheme, recently
proposed by Filip et al.,36,38
involving strongly forced waves
at frequency shifts well below resonance in a marginally un-
derdense plasma, the autoresonant/APTR PBWA enjoys a
number of advantages, in terms of plasma wave amplitude,
robustness, and quality. In previous sections we have seen
how, for given drive laser intensity, autoresonant excitation
yields longitudinal fields that can be considerably higher
than the RL limit set by relativistic detuning of the plasma
FIG. 4. Plasma wave excitation for a 800 nm Ti:sapphire laser with /p=25, intensity 21017 W/cm2 =0.09. APTR parameters are 0=1.2, and =1 at t 1.4 ps =0.00225. Total chirp is 3% of the laserfrequency. In a, we see that the DMG and APTR scheme give approxi-
mately the same final fields, about three times that for on-resonance. In b,we changed p by 10%, and the excitation for the DMG and the on-resonantscheme drop considerably, while APTR maintains its large excitation. Panel
c indicates the insensitivity of APTR to errors of 35% in the plasmadensity, while still yielding large amplitude plasma waves.
FIG. 5. Fluid simulations of the Ti:sapphire example from Sec. V B. The
solid line in a demonstrates the RL oscillations in Ez characteristic ofresonant excitation, while the dotted line indicates saturation of the DMG
chirp from resonance due to failed frequency locking. The solid line in bshows autoresonant excitation of the plasma wave to amplitudes above E0.
The dotted line in b indicates some nontrivial evolution of the lasers;
nevertheless, overall agreement with the simplified QSA model is quitegood.
123103-10 Lindberg et al. Phys. Plasmas 13, 123103 2006
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
11/12
wave. We have also seen how APTR, i.e., slowly sweeping
the frequency downward through resonance, provides a
much greater degree of robustness to density mismatches,
since neither the final amplitude nor frequency of the plasma
wave is very sensitive to the precise location of the actual
linear resonance.
While our simplified model has demonstrated the robust-
ness of autoresonant excitation with respect to global density
mismatches and to small changes in the laser parameters, webelieve that some of this robustness should persist in the
presence of moderate spatio-temporal variations and nonuni-
formity in either the plasma or the laser dynamics. While
other excitation schemes will yield highly irregular acceler-
ating structure in the presence of such variations, our au-
toresonant approach enjoys an intrinsic insensitivity and per-
sistence, due to the local nature of the phase-locking.
Provided only that the magnitude and scales for the nonuni-
formity are such as to allow an eikonal treatment of the
waves, we expect24
that at each position, the local plasma
response will be autoresonantly excited by the drive, based
on the local, slowly-varying values of the plasma frequency,
drive amplitude, beat frequency, and chirp rate. Not all spa-
tial regions will reach precisely the same final amplitude, but
the wave is excited more or less everywhere until local satu-
ration, so the final variations should be considerably less
than in standard approaches. Although plausible, given what
has been demonstrated in previous analyses of autoresonant
phenomena, this expectation should also be verified in more
realistic simulations.
The suitability of the excited plasma waves for relativis-
tic particle acceleration depends not only on the magnitude
of the peak electric fields, but even more crucially on the
uniformity of the phase and phase velocity, and on the degree
of phase-locking to the external drive. With variations in thegroup velocity vg expected to be small vg varies appreciablyonly after propagation lengths of order the aforementioned
dispersion length Ldisp 3/p
3c/p, phase coherence ofthe plasma wave will depend on how closely vp follows the
essentially constant vg of the laser. Particle-in cell PICsimulations of Filip et al.
38suggest that for the RL/TD
scheme, the effective phase velocity of the nonlinear plasma
wave can vary appreciably, i.e., 10% to 20%, reflecting phase
slippage of the Langmuir wave primarily as a result of rela-
tivistic detuning, while the plasma wave produced in their
nonresonant PBWA scheme exhibits substantially less phase
slippage. Since our autoresonant scheme also yields
frequency-locked excitation, we expect to find an accelerat-ing field of uniform phase that is everywhere directly related
to the local phase of the driving beat wave.
This ostensible ability to phase-lock the plasma wave to
the beat-wave of the applied drive lasers is an appealing
feature of both the nonresonant PBWA and the autoresonant
PBWA, since the timing of electron injection, whether based
on an external cathode39
or internal optical method,40,41
will
be provided by presumed knowledge of the laser phase. Be-
cause of its potential importance, this phase-locking deserves
careful investigation. But first, one must distinguish phase-
locking from frequency-locking, however much these terms
are conflated. Perfect phase-locking implies perfect
frequency-locking, and conversely, at least up to a constantphase, but in the case of imperfect entrainment, it is possibleto achieve good frequency-locking without adequate phase-
locking, or the converse. Whether the relative error in phase-
matching or in frequency-matching between driving and
driven oscillation is greater depends on whether the Fourier
content of fluctuations in the phase is primarily at higher or
lower frequencies than the drive frequency itself.
For the purpose of matched particle injection, it is phase-
locking that is desired, yet some caution is warranted in
claims of true phase-locking in either the nonresonant or au-toresonant PBWA. In any frequency-locked PBWA scheme,
the nonlinear frequency of the Langmuir wave may be
closely entrained to the precisely known drive frequency, but
this does not necessarily imply that the absolute phase of the
Langmuir wave may be precisely known. As a mechanism
for phase-locking in the nonresonant PBWA, the authors ap-
peal to the claim that an harmonic oscillator, strongly driven
off resonance, remains synchronous with the driving force.
However, this intuition holds only for damped linear oscilla-
tors after the transient is allowed to decay. If a linear oscil-
lator with natural frequency n is forced by a sinusoidal
drive with frequency dn, then, independent of the
FIG. 6. Evolving phase difference between the maxima of the laser beat and
the longitudinal electric field Ez which may differ from in the text by aconstant. Plot a uses the CO2 laser parameters of Fig. 3, which demon-strates nearly constant phase locking throughout autoresonant excitation for
40 t100, permitting good control of electron injection. Plot b uses theTi:sapphire parameters of Fig. 4 and shows tight phase-locking during the
beginning of excitation t2 ps, at which time Ez 225 GV/m.
123103-11 Autoresonant beat-wave generation Phys. Plasmas 13, 123103 2006
Downloaded 08 Dec 2006 to 132.64.1.37. Redistribution subject to AIP license or copyright, see http://pop.aip.org/pop/copyright.jsp
8/3/2019 R. R. Lindberg et al- Autoresonant beat-wave generation
12/12
strength of the drive, the driven oscillator will exhibit persis-
tent, oscillating variations in phase relative to the drive phase
d=dt+d0, which are Od/n; i.e., only first orderin their frequency ratio. There is no obvious reason to expect
in general that nonlinearities will improve matters.
In the autoresonant case, we encounter good frequency-
locking with comparably good phase locking only over some
portion of the Langmuir excitation, depending on the drive
parameters. In Fig. 6 we plot, for our experimentally moti-vated parameters of Sec. V, the phase lag between the beat-
wave of the lasers and the plasma wave, numerically esti-
mated by the offset between corresponding relative maxima.
In the case of the CO2 laser parameters, Fig. 6a indicatesthat once the laser beat passes through the linear resonance,
the plasma wave phase tightly locks with the drive phase and
remains so until adiabaticity is lost and the excitation satu-
rates. In this case, a phase-locked injection scheme could
reliably place electron bunches near the maximum accelera-
tion gradient 11 GV/m.While the Ti:sapphire example in Fig. 6b also initially
experiences tight phase-locking to the drive, we see that the
phase difference between the laser beat and the Langmuirwave drifts during the latter stages of excitation, eventually
reaching at saturation. Thus, the highest-amplitude fields
during the final stages of excitation experience some phase
slippage from the lasers, and one may not be able to reliably
inject particles into the largest accelerating gradients
achieved. Nevertheless, the longitudinal field reaches an am-
plitude of approximately 225 GV/m while the phase is
tightly locked, thus permitting controlled, phase-locked in-
jection into accelerating fields above both ERL 125 GV/mand E0 200 GV/ m.
VIII. CONCLUSION
We have introduced a straightforward, seemingly minor,
modification of the DMG scheme for the chirped-pulse
PBWA, based on the nonlinear phenomenon of autoreso-
nance with adiabatic passage through resonance, which nev-
ertheless enjoys certain advantages over previous ap-
proaches. Rather than starting at the linear resonance and
chirping downward using a profile expected to match the
decreasing nonlinear plasma frequency, we start above reso-
nance and sweep the beat frequency downward through the
resonance, at any sufficiently slow chirp rate, such that the
plasma wave frequency automatically self-locks to the drive
frequency, and the plasma wave amplitude automatically ad-
justs itself consistent with this frequency. This autoresonant
excitation achieves higher plasma wave amplitudes at mod-
erate laser intensities, and, most importantly, appears to be
much more robust to inevitable uncertainties and variations
in plasma and laser parameters. Preliminary analysis has
been performed within a simplified analytic and numerical
model, and wake excitation has been studied using realistic
parameters for Ti:sapphire and CO2 laser systems. The re-
sults have been substantiated with self-consistent 1D fluid
simulations, and warrant extending investigation to higher-
dimensional geometries and more realistic plasma inhomo-
geneities via numerical solution with additional fluid and
PIC codes.
ACKNOWLEDGMENTS
This research was supported by the Division of High
Energy Physics, U.S. Department of Energy, Grant Number
DE-FG02-04ER41289, and by the USIsrael Binational Sci-ence Foundation Grant No. 2004033.
1J. Tajima and J. Dawson, Phys. Rev. Lett. 43, 267 1979.
2N. Kroll, A. Ron, and N. Rostoker, Phys. Rev. Lett. 13, 83 1964.
3B. Cohen, A. Kaufman, and K. Watson, Phys. Rev. Lett. 29, 581 1972.
4M. Rosenbluth and C. Liu, Phys. Rev. Lett. 29, 701 1972.
5C. Joshi, W. Mori, T. Katsouleas, J. Dawson, J. Kindel, and D. Forslund,
Nature 311, 525 1984.6
B. Amini and F. Chen, Phys. Rev. Lett. 53, 1441 1984.7
C. Tang, P. Sprangle, and R. Sudan, Phys. Fluids 28, 1974 1985.8
C. Clayton, C. Joshi, C. Darrow, and D. Umstadter, Phys. Rev. Lett. 54,
2343 1985.9
R. Noble, Phys. Rev. A 32, 460 1985.10
W. Mori, IEEE Trans. Plasma Sci. PS-15, 88 1987.11
P. Mora, D. Pesme, A. Hron, G. Laval, and N. Silvestre, Phys. Rev. Lett.61, 1611 1988.
12Y. Kitigawa, T. Matsumoto, T. Minamihata et al., Phys. Rev. Lett. 68, 48
1992.13
M. Everett, A. Lal, D. Gordon, C. Clayton, K. Marsh, and C. Joshi, Nature
368, 527 1994.14
C. Clayton, C. Joshi, K. Marsh, C. Pellegrini, and J. Rosenzwieg, Nucl.
Instrum. Methods Phys. Res. A 410, 378 1998.15
E. Esarey, P. Sprangle, J. Krall, and A. Ting, IEEE Trans. Plasma Sci. 24,
252 1996.16
A. Akhiezer and R. Polovin, Sov. Phys. JETP 3, 696 1956.17
J. Dawson, Phys. Rev. 113, 383 1959.18
C. McKinstrie and D. Forslund, Phys. Fluids 30, 904 1987.19
G. Shvets, Phys. Rev. Lett. 93, 195004 2004.20
J. Matte, F. Martin, N. Ebrahim, P. Brodeur, and H. Pepin, IEEE Trans.
Plasma Sci. PS-15, 173 1987.21
M. Deutsch, B. Meerson, and J. Golub, Phys. Fluids B 3, 1773 1991.22A. Loeb and L. Friedland, Phys. Rev. A 33, 1828 1986.23
B. Meerson and L. Friedland, Phys. Rev. A 41, 5233 1990.24
L. Friedland, Phys. Fluids B 4, 3199 1992.25
L. Friedland, Phys. Rev. E 58, 3865 1998.26
P. Sprangle, E. Esarey, and A. Ting, Phys. Rev. Lett. 64, 2011 1990.27
D. Forslund, J. Kindel, and E. Lindman, Phys. Rev. Lett. 30, 739 1973.28
C. Max, J. Arons, and A. Langdon, Phys. Rev. Lett. 33, 209 1974.29
P. Sprangle, E. Esarey, and A. Ting, Phys. Rev. Lett. 64, 2011 1991.30
Table of Integrals, Series, and Products, edited by I. Gradshteyn and I.
Ryzhik Academic, New York, 1980.31
B. Chirikov, Phys. Rep. 52, 263 1979.32
F. Lewis, Trans. ASME 54, 253 1932.33
E. Grosfeld and L. Friedland, Phys. Rev. E 65, 046230 2002.34
J. Fajans, E. Gilson, and L. Friedland, Phys. Plasmas 6, 4497 1999.35
D. Strickland and G. Mourou, J. Opt. Commun. 56, 219 1988.36
C. Filip, S. Tochitsky, R. Narang, C. Clayton, K. Marsh, and C. Joshi, AIP
Conf. Proc. 647, 770 2002.37
B. Shadwick, G. Tarkenton, E. Esarey, and W. Leemans, IEEE Trans.
Plasma Sci. 30, 38 2002.38
C. Filip, R. Narang, S. Tochitsky, C. Clayton, P. Musumeci, R. Yoder, K.
Marsh, J. Rosenzweig, C. Pellegrini, and C. Joshi, Phys. Rev. E 69,
026404 2004.39
C. Clayton and L. Serafini, IEEE Trans. Plasma Sci. 24, 400 1996.40
E. Esarey, R. Hubbard, W. Leemans, A. Ting, and P. Sprangle, Phys. Rev.
Lett. 79, 2682 1997.41
C. Schroeder, P. Lee, J. Wurtele, E. Esarey, and W. Leemans, Phys. Rev. E
59, 6037 1999.
123103-12 Lindberg et al. Phys. Plasmas 13, 123103 2006
Top Related